SUMMARY: Environment issue, long $CLASSPATH and such.

From: Mike Barnett <barnettmike_at_attbi.com>
Date: Sat Mar 09 2002 - 16:05:54 EST
Original posting follows this summary.

Summary: The general consensus is that I would get more value out of tuning
the kernel than worrying about long pathnames in the environment
definitions.
http://sunsolve.sun.com/private-cgi/retrieve.pl?type=2&doc=stb/1442
Thanks, Karl Vogel, for that link.

There were other suggestions which taught me some new and interesting
things.

Eshafto wrote,
Well, it might take some careful benchmarking to detect any difference,
but if you were absolutely obsessed with squeezing out performance, why
not:

create a RAM-based filesystem, say /classes
for i in `echo $REAL_CLASSPATH | sed "s/:/ /g" `; do
cp $i/*  /classes
done
Set www's CLASSPATH to /classes.

Run the for loop whenever you change one of the class files. Of course,
to get 100% compatibility, you'd have to copy them in reverse order so
that the first ones would overwrite the last. But if you've got unique
names for all your class files, that won't matter.

Here is the original posting.
--- Mike Barnett <barnettmike@attbi.com> wrote:
> Hello Solaris Managers,
>
> I am trying to improve performance on a large and heavily used Solaris 8
> system web based application.
> Amongst the many other items I need to look at, I was curious what your
> opinions might be on extremely long environment variable definitions.
> In example, if a $CLASSPATH has over 70 specific paths, well over 2000
> characters, would there be negligible performance degradation for me to be
> concerned about if we are getting a couple thousand hits per minute or
> more?
> I want to propose creating a single directory and simply soft linking all
> these targets into this one area. I can't help but wonder about inode
> lookups and disk I/O regarding this long path. And I noticed there are a
few
> other details like the before mentioned $CLASSPATH in the configuration
file
> used to run the application.
> Your opinion would be greatly appreciated.
>
> I thank you in advance.
>
> Mike Barnett
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
Received on Sat Mar 9 15:02:12 2002

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:42:36 EST